Researches On Error Control Methods In Scalable Video Coding

Scalable Video Coding(SVC),which is the scalable extension of H.264/AVC, is currently a powerful solution to video applications over heterogeneous networks and diversified end-users.SVC can provide one single embedded bit-stream,from which several sub-bit-streams can be extracted,to meet the constraint of transmission rates and the requirements of end-users in spatial,temporal,and SNR (Signal-to-Noise Ratio) quality.Though SVC makes full use of redundancy in space, time,and inter-layer to improve coding efficiency,for error-prone channel(e.g., wireless communication networks),channel errors can not only damage current frame,but also propagate to subsequent frames,such that video quality is degraded drastically.Error control is an efficient tool to stop degradation of subjective and objective quality caused by channel errors in error-prone channel,and can be implemented in encoder,decoder,or transmission channel of video communication system.It is a promising technique in theory and practice to design error control methods for SVC.On the basis of SVC framework,this dissertation brings forward various error control techniques.The main contents and novelties are as follows:1) This dissertation proposes error concealment in decoder.According to the characteristics of SVC,this dissertation makes full use of spatial,temporal,and inter-layer information received correctly at decoder to perform intra-layer error concealment and inter-layer error concealment.In addition, OBMC(Overlapped Block Motion Compensation) inter-layer error concealment is utilized to cope with packet losses in enhancement layer.2) This dissertation proposes error resilient coding in encoder.Error resilient coding proposed in this dissertation injects redundancy into the scalable bit-stream to help decoder in recovery from potential channel errors and in improvement of rate-distortion performance.According to inter-layer characteristics of SVC,LA-RDO(Loss-Aware Rate-Distortion Optimization) algorithm,which was proposed by the author and adopted by JVT(Joint Video Team),can be applied in an intra-refresh style.Furthermore,inter-layer error concealment optimization at encoder is applied in the scenarios where packet losses happen in enhancement layer.3) This dissertation proposes error resilient transcoding in transport-level. The proposed error resilient transcoding is a transport-level error control method.It is applied over heterogeneous networks,to optimize error resilient capability according to the feedback information of packet losses at decoder.This work was first brought forward in SVC,and it can not only enhance error resilient capability of SVC bit-stream,but also maintain consistent bit-rates between input and output of the transcoder.4) This dissertation proposes discardable data adaptation in encoder.The proposed discardable data adaptation method uses data partitioning and slice group with error resilient capability,to separate base layer picture into discardable and non-discardable types.When diacardable data are totally discarded, rate-distortion performance in unicasting enhancement layer video can be improved, meanwhile the bit-stream can be decoded correctly.Compared with single layer methods,besides temporal and spatial information, inter-layer information can also be employed to design error control techniques for SVC.Although many error control methods in single layer have been brought forward,they cannot be directly applied to SVC.In this dissertation,thorough researches are performed in SVC,and several error control methods are proposed, which significantly improve error control capability of SVC bit-stream.